Method and apparatus for resource allocation in wireless communication system
View Patent ↗An operation method in a communication system may comprise: obtaining information on a per-beam required traffic amount; determining whether the per-beam required traffic amount can be serviced while satisfying a first condition according to a first model generated through pre-training in a first machine learning structure; in response to determining that the per-beam required traffic amount can be serviced while satisfying the first condition, calculating per-beam bandwidth allocation information based on the per-beam required traffic amount; calculating per-beam power allocation information based on the per-beam bandwidth allocation information; identifying whether an SNR condition included in the first condition is satisfied based on the per-beam power allocation information; and in response to identifying that the SNR condition is satisfied, outputting the per-beam bandwidth allocation information and the per-beam power allocation information.
1. An operation method of a first apparatus in a communication system, the operation method comprising:
obtaining information on a per-beam required traffic amount;
determining whether the per-beam required traffic amount can be serviced while satisfying a first condition including an available total bandwidth condition and an available total power condition according to a first model generated through pre-training in a first machine learning structure;
in response to determining that the per-beam required traffic amount can be serviced while satisfying the first condition, calculating per-beam bandwidth allocation information based on the per-beam required traffic amount;
calculating per-beam power allocation information based on the per-beam bandwidth allocation information;
identifying whether a signal-to-noise ratio (SNR) condition included in the first condition is satisfied based on the per-beam power allocation information; and
in response to identifying that the SNR condition is satisfied, outputting the per-beam bandwidth allocation information and the per-beam power allocation information.
2. The operation method according to claim 1 , wherein the determining comprises:
generating first input data by converting the information on the per-beam required traffic amount into a vector;
inputting the first input data to the first model; and
identifying an output value of the first model.
3. The operation method according to claim 2 , wherein the first machine learning structure has a perceptron structure, the output value of the first model, which has a positive value, means that the per-beam required traffic amount can be serviced while satisfying the first condition, and the output value of the first model, which has a negative value, means that the per-beam required traffic amount cannot be serviced while satisfying the first condition.
4. The operation method according to claim 2 , wherein the pre-training in the first machine learning structure is performed based on a second model after the second model is generated through pre-training in a second machine learning structure for calculation of the per-beam bandwidth allocation information.
5. The operation method according to claim 1 , wherein the calculating of the per-beam bandwidth allocation information comprises:
inputting first input data generated based on the information on the per-beam required traffic amount into a second model generated through pre-training in a second machine learning structure; and
obtaining output data output from the second model,
wherein the output data includes the per-beam bandwidth allocation information.
6. The operation method according to claim 5 , wherein the second machine learning structure has a machine learning structure according to a linear regression learning scheme, and the pre-training in the second machine learning structure is performed based on an exhaustive search scheme in a direction in which a value of a loss function calculated based on first bandwidth allocation information output based on information on a first required traffic amount is minimized.
7. The operation method according to claim 1 , further comprising, after the determining,
in response to determining that the per-beam required traffic amount cannot be serviced while satisfying the first condition, performing an affine projection operation for calculating a reduced required traffic amount reduced from the per-beam required traffic amount; and
calculating the per-beam bandwidth allocation information based on second input data generated as a result of the affine projection operation and including information on the reduced required traffic amount.
8. The operation method according to claim 7 , further comprising, after the identifying,
in response to identifying that the SNR condition is not satisfied, adjusting a boundary value used in the affine projection operation;
performing the affine projection operation based on the adjusted boundary value; and
calculating the per-beam bandwidth allocation information based on third input data generated as a result of the affine projection operation performed based on the adjusted boundary value.
9. A first apparatus in a communication system, comprising:
a processor;
wherein the processor causes the first apparatus to:
obtain information on a per-beam required traffic amount;
determine whether the per-beam required traffic amount can be serviced while satisfying a first condition including an available total bandwidth condition and an available total power condition according to a first model generated through pre-training in a first machine learning structure;
in response to determining that the per-beam required traffic amount can be serviced while satisfying the first condition, calculate per-beam bandwidth allocation information based on the per-beam required traffic amount;
calculate per-beam power allocation information based on the per-beam bandwidth allocation information;
identify whether a signal-to-noise ratio (SNR) condition included in the first condition is satisfied based on the per-beam power allocation information; and
in response to identifying that the SNR condition is satisfied, output the per-beam bandwidth allocation information and the per-beam power allocation information.
10. The first device according to claim 9 , wherein in the determining, the processor further causes the first apparatus to:
generate first input data by converting the information on the per-beam required traffic amount into a vector;
input the first input data to the first model; and
identify an output value of the first model.
11. The first device according to claim 10 , wherein the first machine learning structure has a perceptron structure, the output value of the first model, which has a positive value, means that the per-beam required traffic amount can be serviced while satisfying the first condition, the output value of the first model, which has a negative value, means that the per-beam required traffic amount cannot be serviced while satisfying the first condition, and the pre-training in the first machine learning structure is performed based on a second model after the second model is generated through pre-training in a second machine learning structure for calculation of the per-beam bandwidth allocation information.
12. The first device according to claim 9 , wherein in the calculating of the per-beam bandwidth allocation information, the processor further causes the first apparatus to:
input first input data generated based on the information on the per-beam required traffic amount into a second model generated through pre-training in a second machine learning structure; and
obtain output data output from the second model,
wherein the output data includes the per-beam bandwidth allocation information.
13. The first device according to claim 12 , wherein the second machine learning structure has a machine learning structure according to a linear regression learning scheme, and the pre-training in the second machine learning structure is performed based on an exhaustive search scheme in a direction in which a value of a loss function calculated based on first bandwidth allocation information output based on information on a first required traffic amount is minimized.
14. The first device according to claim 9 , wherein the processor further causes the first apparatus to, after the determining,
in response to determining that the per-beam required traffic amount cannot be serviced while satisfying the first condition, perform an affine projection operation for calculating a reduced required traffic amount reduced from the per-beam required traffic amount; and
calculate the per-beam bandwidth allocation information based on second input data generated as a result of the affine projection operation.
15. The first device according to claim 14 , wherein the processor further causes the first apparatus to, after the identifying,
in response to identifying that the SNR condition is not satisfied, adjust a boundary value used in the affine projection operation;
perform the affine projection operation based on the adjusted boundary value; and
calculate the per-beam bandwidth allocation information based on third input data generated as a result of the affine projection operation performed based on the adjusted boundary value.